The Research of Neural Network and EMD Methods Used in the Prediction of Machine Running State

Abstract:

The paper puts forward a method of ventilator running state prediction which is based on EMD and Neural Network, which aims at ventilators non-stationary vibration signals. First of all, vibration data are decomposed into eight intrinsic mode signals and one residual quantity by EMD. Then Neural Network is adopted to predict the decomposition components and reconstruct them. Finally, the final prediction results are achieved. In this paper, the ventilator vibration signals of Tangshan Iron and Steel Group Coking Plant from March 1, 2012 to July 1, 2012 are used as the research data. Comparative study method is adopted to compare this prediction method with direct Neural Network prediction in terms of results, which shows that this method has better accuracy.

• Periodical: Advanced Materials Research (Volumes 753-755)
• Pages: 2183-2186
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.753-755.2183
• Online since: August 2013
• Authors: Shi Man Xie, Xiao Le Wang, Zhao Yan Xuan, Pei Cai, Fei Yang
• Keywords: Empirical Mode Decomposition (EMD), Neural Network (NN), Running State Prediction

References

[1] Jiangping Wang. Mechanical equipment fault diagnosis technology and application [M]. Xi'an: The Northwest industry university presses, 2011: 180-182. In Chinese.

[2] Changhong Dong. Neural network and Applications[M]. Beijing: National Defense of Industry Press, 2005: 124-143. In Chinese.

[3] Jian Pei, Jiawei Han, Mortazavi-Asl,B. PrefixSpan: mining sequential patterns efficiently by prefix-projected pattern growth [J]. Data Engineering, 2001, 8(4): 215-224.

DOI: 10.1109/icde.2001.914830